AI Use Notice

Version 1.11.0

AI Use Notice — version 1.11.0 1. You are talking to an AI Karina AI is an artificial intelligence system. It is not a person and not a doctor. Answers are generated automatically. 2. How answers are produced Part of the answers the app produces itself, without turning to a model: the water and calorie targets, the data in your own logs, reference values — that is computed on your device. Everything else is generated by a model. The information you chose to provide (questionnaire, lab results, journal) is used only to personalize explanations. The author’s library of materials (the books, guides and videos in the app’s tab) is a separate feature: you open and read those files yourself, and they take no part in producing chat answers. Chat answers are generated by a third-party model. The primary one is GPT-5 nano on the Microsoft Azure platform: the questions you write to Karina AI, health topics included, are passed to it, and it runs in Microsoft’s European data zone (Sweden). The model was developed by OpenAI, but it is hosted and run by Microsoft itself — the model’s developer does not receive your questions. If the primary model is busy or does not answer, the question goes to the backup model on the same platform — DeepSeek-V4-Flash; it was developed by DeepSeek but is hosted and run by the same Microsoft in the same European zone. We may replace the specific model with another one hosted on the same platform: the recipient, the processing region and the rules in this section stay the same. If it does not answer, the same question goes to Google Gemini, and if that fails too, to Anthropic Claude — both of those work outside the EEA. Google Gemini is also the model that reads the lab report you upload, recognizes food from a photo and counts calories from your voice (paid tier: the models are not trained on your data). Anthropic never receives reports. Separately, something easy to miss: Microsoft watches that the model is not used to do harm, and if a violation is suspected, a sample of questions and answers may be stored and read by its authorized employees — inside the European Economic Area. The details are in the Privacy Policy, section 4. Along with a new question, several of the most recent turns of your chat with Karina AI are passed to the provider — up to ten. Without them an answer would lose the thread: you write “so what do I do about it?”, and the model no longer knows what was being discussed. This chat history is not kept on our server — it is stored on your phone, and it goes to the provider only for the time it takes to produce the answer. Your chat with Karina AI is not the same thing as your personal messages to Karina herself. The chat is a conversation with a model; the personal messages are a conversation with a person, they are kept on our server and are described in the Privacy Policy (sections 2, 5 and 8). This notice is only about the chat with the AI. 3. What the AI does here and what it does not do This is not a medical app. For lab results we display what is printed on your report and keep it in digital form. The report is read by an AI model — Google Gemini, the one named in section 2: it transfers the marker name, the value, the unit, the reference range and the date from it — and it can also get them wrong, so check what was transferred against the original. The reference range is taken from your report, the one your laboratory stated: the app does not substitute a range of its own and does not recalculate your results against its own tables. The “in range / out of range” label is the result of comparing your value with the range from your own report; if there is no range there, the marker is labeled “not evaluated”. Karina explains what a marker means and what such values usually indicate — informationally and as common knowledge, not as a conclusion about your health and not as the reason for your own deviation. Karina does NOT do the following: • she does not make diagnoses and does not assert that you have a disease; • she does not prescribe or discontinue treatment, and does not select medications or dosages; • she does not replace your doctor: the final conclusion in your case is made by them. 4. Limits of the AI AI can be wrong, including when reading numbers off a report. Check values against the original. The answers are educational in nature and do not replace a consultation with a doctor. 5. When to see a doctor Any interpretation of lab results, changes in how you feel, starting and stopping medications and supplements — discuss with a qualified medical professional. 6. Emergencies Chest pain, difficulty breathing, signs of a stroke, a severe allergic reaction, loss of consciousness, heavy bleeding, an overdose, thoughts of harming yourself — contact emergency services immediately. 7. Reporting a problematic answer If an answer from Karina AI seems unsafe or wrong, tell us: info@karinamore.com. The history of your chat with Karina AI is stored on your device — you need it yourself, and it lets you show us a problematic answer. We do not keep it on our side; that part of the most recent turns goes to the model provider along with a new question is written in section 2. Your personal messages to Karina are a different thing and they do live on our server — see the Privacy Policy.